Objects which are preceptually similar are often difficult to discriminate from each other and are often classified together. Objects that differ considerably in similarity are relatively easy to discriminate and may be separately classified. The objectives of this research are to test whether these generalizations available from the study of perceptual patterns used in the laboratory apply also to patterns used in the classroom, such as letters and words. Letters of the normal alphabet, like b and d, are frequently confused and, for some children, are difficult to learn as unique forms. An alphabet font is proposed in which such letters are slightly modified to make them perceptually distinct. We propose these distinctive letters will be more readily learned by young children than are normal letters. Also, since the similarity between normal letters and their altered counterparts has simultaneously been kept great, children may have little or no difficulty transferring between alphabets. Some completed pilot work is reported which supports these hypotheses. The proposed work has two major components. The first is to test if children can learn the alphabet, and learn to read, more readily if they are trained initially with the distinctive letters rather than normal letters. The second component of our work is to experimentally contrast predictions from several feature theories of pattern perception with a previously proposed (Lockhead, 1970) similarity theory.